Hey there fellas.
Since people sometimes make suggestions about LoRas and ask if/when they will come out, I decided to create this article with the Loras I have planned, and I'll try to briefly explain my reasoning when planning/creating LoRas.
The LoRas I have planned/currently working on:
Currently working on:
Sexercise (Dataset prep)
Planned (in no particular order):
Herbal Love Story
Collapse & Rewind
I Need You Noona
Introduction To An Affait
My Kingdom/Silent War
Is there an Empty Room
What Do I Do Now?
What do you Take me for?
The Runaway Family
My reasoning behind my releases:
This is a Hobby thing, and I have a fulltime job: this one is very self explanatory, but might as well get it out of the way.
My time for A.I is very limited: I have in average 6hrs from finishing work until I sleep. Let's say that at most, 3h of these are dedicated to A.I art. This includes checking out what the community is creating, reviewing models, engaging with other users, and working on my models. As you can imagine, this limits my output significantly, but I think is important not to focus exclusively on output, as the community has so much to teach.
I have my preferences: If you've seen my models at all, you may have noticed at pattern. To be blunt: I like mature/adult characters, and I despise high schoolers/lolis/juvenile characters. So if you suggest characters that fit those categories, I'll look it up, but will not create any models for it.
Models based on screenshots are easier: For the majority of cases, the amount of work to produce a Lora based on a animation is a fraction of what is necessary to create models of Manhwa. I just need to get the keyframes, sometimes some fanart to get full body images, tag, and train. I can do this in less than 3 hours. For manhwa, I need to save the images for each chapter, then open then in photoshop to crop, sometimes I can't find the characters names on databases so I need to read some of the manhwa to see their names and age, and worst case scenario, I need to remove fx, text, and speech bubbles using photoshop. The dataset prep alone can take more than 3hours sometimes, if the character appears less frequently. So the output is waaay slower. For this reason, I'll sometimes pump models based on animations while I work the manhwa characters on the side. This might be a skill issue, so let me know if you have better ideas on how to make data prep for manhwas.
I get easily distracted: I have ADHD and it's very hard to focus on a task that takes too long, and as I described on point 4, manhwa models are exactly that. So I have at any given time, several manhwas I'm going through at the same time. So I'll sometimes dump a bunch of models in sequence, and then stay silent for a long period of time. The models based on animations are a way of padding this silent period with something for the community.
It's hard to find good male characters: I don't have a problem with creating models for male characters, however, I find it hard to justify creating a model for most of the male characters I see. Most of them look boring/generic AF, or are straight up huge assholes.
That's all I can think for now, about my decision making. Keep in mind this is not a rant, I'm just trying to be as transparent as possible.
Non-related: CivitAI will very often not notify me of new reviews of my models, not sure why. So if I don't comment on your review, or don't reply to your comment, it's nothing personal. I might just not have seen it.
Feel free to ask anything I missed.